Alerts

Alerts come from three on-server tools: Gatus (is each service answering), Healthchecks (did each scheduled job report in) and Beszel (are CPU, memory and disk healthy). The overview is in Monitoring and alerts. This page covers the settings and the first-time wiring.

Fields

Settings > Alerts and missed-job reporting holds the off-site heartbeat address. Scheduled jobs report to the Healthchecks on the server, which raises the alarm when a report is late or failed.

Field	Purpose
Off-site heartbeat address	The ping address of a check at an outside monitoring service, such as a free healthchecks.io check. Used by the off-site heartbeat (Catena Pro), which you turn on in Schedules.

A monitor on the server cannot report an outage that takes the whole server down. The off-site heartbeat closes that gap: on its schedule, the server checks that its status page, job monitor and resource monitor answer, then calls the off-site address. A plain call means all is well; a failure call names the monitors that stopped answering; no call at all means the server is down, and the outside service raises the alert. Saving the address alone sends nothing: the heartbeat runs once you turn it on in Schedules, where its section shows the period and grace time to set on the outside check. When your subscription ends, the calls stop, so pause the outside check then.

Alerts also reach your admin email, through the service chosen in Outgoing email: while outgoing mail is set, the server keeps an email channel to your admin address in Healthchecks, attached to every check when the channel is created. A channel detached from a check in Healthchecks stays detached. With No outgoing mail, the channel is removed and alerts reach only the channels you add in Healthchecks.

The section ends with Save and apply; the services that use these values restart briefly.

How alerts flow

1. Gatus probes the shared services (the admin panel, Healthchecks, Keycloak, Beszel, the Cloudflare tunnel and the reverse proxy) and every routed application on a fixed interval (about one to two minutes). After 3 consecutive failures it raises an alert, and after 2 successes it resolves it.
2. Each Gatus alert is posted to Healthchecks as one check per service, named gatus-<service>, created on the first failure.
3. Healthchecks sends each alert to your admin email and to the channels added in its own interface.
4. Scheduled jobs (backup, offsite copy, nightly maintenance, restart probe, off-site heartbeat) ping Healthchecks as they run. A missed ping is itself the alert. The checks of the backup, the offsite copy, the nightly maintenance (Nightly maintenance, told when a night starts, fails or ends) and the off-site heartbeat follow the schedules of the jobs that report to them; while none of those jobs is turned on, the check is paused, so it never goes late. On every edition, the vulnerability watch also creates one check per vulnerability it alerts on (see Security alerts). The nightly maintenance reports to the backup and offsite copy checks only while its backup and offsite copy steps are on (see Choosing the steps).
5. Beszel watches the server with rules created once and never overwritten: status (agent silent) for 5 minutes, CPU at 90% for 10 minutes, memory at 90% for 10 minutes, disk at 85% for 10 minutes. Its alerts go through Healthchecks, so they reach the same channels.

Send error reports to Catena support

Settings > Error reports to Catena support lets a server on Catena Pro or Catena Business tell Catena support when one of Catena’s own checks fails or recovers, so support can help you before a small fault becomes a large one. It is off until you turn it on.

To turn it on:

1. Read the report the section shows. It is the report this server would send right now.
2. Tick the box that confirms you agree, then press Turn error reports on. Within five minutes the server signs up with Catena support.

To turn it off, press Turn error reports off. The server stops sending at once and signs off.

A report names the check, whether it is failing, since when, the application it concerns, your Catena version and your edition. It never holds what is inside your applications: no files, log lines, user accounts, email or network addresses, or domain names. A report is sent only when the list of failing checks changes or a one-off event happens.

Below the switch, the section lists the last 100 reports sent, newest first, each with the exact text that left the server and whether Catena support received it.

When a failed job wrote text about what went wrong, that text stays on your server. The list offers it for each report that has some:

• Download the job output saves the text as it is.
• Review and send to Catena support opens the text in a box. Remove whatever you do not want support to see, then press Send this text to Catena support. Only the text in the box is sent, and the list shows exactly what you sent.

Support answers Monday to Friday, 08:00 to 17:00. Error reports are not round-the-clock monitoring, and a server that is switched off or cannot reach the internet sends nothing: for that, use the off-site heartbeat above.

Add notification channels in Healthchecks

Alerts reach your admin email with no step here. You add more channels, for other people or other tools, in Healthchecks:

1. Open https://healthchecks.yourdomain.com and sign in (single sign-on).
2. Open the project’s integrations page and add a channel: email, Slack, ntfy, a webhook, or any other the page offers.
3. Check that the channel is attached to the checks you want notified. New checks, including the gatus- ones, use the project’s channels.

Email channels need Outgoing email.

Where to look

• https://gatus.yourdomain.com: the Gatus status page.
• https://healthchecks.yourdomain.com: Healthchecks, with each job’s last ping.
• https://beszel.yourdomain.com: Beszel, with the server’s resource graphs.
• The System page of the admin panel shows active alerts and jobs not reporting yet.

You can change the subdomain names in Domain and Cloudflare.

Troubleshooting

• A job shows late or down in Healthchecks: the job failed, or did not run at its schedule. Check the Log and Schedules.
• No alert email arrived: Outgoing mail is set to No outgoing mail, your admin email channel was detached from that check in Healthchecks, or the mail went to the spam folder.
• No notification on another channel: that channel is not attached to the check. Attach it as above.
• A channel you added never delivers: check its address or topic in Healthchecks against the receiving side.